Proceedings of the Annual Meeting of the Cognitive Science Society

Emotions regulate social interactions from early in ontogeny, but are difficult to assess in young children. Previous studies used body posture to measure emotion expressions, employing depth-sensor imaging cameras. Advances in artificial intelligence allow researchers to track posture from existing videos. The reported studies explored the feasibility of OpenPose to capture children's emotional expressions. In Study 1, we analysed posture data from previous studies and found that children's expressed valence was positively related to changes in upper-body expansion whereas expressed arousal was related to overall movement. In Study 2, children (n = 64, 5-10 years) recalled emotional episodes of ‘happiness', ‘sadness', ‘pride', and ‘shame'. There were no effects of specific emotion categories on posture changes, but exploratory analyses revealed that recalling positive emotions yielded greater changes in upper-body expansion compared to negative emotions. Together, these results suggest that the valence and arousal of expressed emotions can be captured using OpenPose.